The proposed project will apply a new method of social network analysis to study youth drinking behavior from adolescence into young adulthood and its determinants across gender and racial/ethnic groups. Peer influence plays a central role in theories of adolescent drinking, and is a primary focus of most adolescent prevention programs. However, limitations in the research on peers and drinking preclude drawing strong conclusions about the role of peer relationships as a contributing factor to youth drinking. Researchers have used social network analysis to identify global friendship structures (cliques, liaisons, isolates) and local structures (dyads) to examine theories of peer influence. However, group detection methods such as the standard NEGOPY fail to examine the building blocks of these peer groups, and have virtually ignored the fact that other kinds of network structures (such as triads) affect children's susceptibility to influence or their development of attitudes and behaviors. We propose to apply a more sophisticated approach using the exponential family of random graph distributions, known as p*, to demonstrate how adolescent friendship structures are composed of dyads, triads, tetrads (i.e., groups of four) and possibly larger entities;and how peer effects such as selection and influence can be examined simultaneously within these structures - features that the current group detection methods do not share. Our project has three specific aims: 1. Determine the extent to which friendship networks influence drinking and drinking influences friendship selection using longitudinal friendship data;and evaluate the long-term impact of peer influences on drinking over a 6-year period. A comparison of the standard group identification method (NEGOPY) and the novel p* method will determine whether the latter perspective provides a different, more sophisticated understanding of the mechanisms associated with social influence and social selection. 2. Investigate how different types of peer influence (e.g., best friend, romantic partner, close friends, friends of friends, whole school) are concurrently and prospectively associated with youth drinking behavior;and explore whether these associations are moderated by key network structure characteristics (i.e., size, density). 3. Identify adolescents who are more resilient versus vulnerable to different types of pro-drinking peer influences (from Aims 1 and 2), focusing on characteristics from four domains: personal factors;school factors;family factors;and neighborhood factors. Longitudinal friendship network data from the National Longitudinal Study of Adolescent will be used;it is the largest study of adolescent friendship networks ever conducted, as well as the only one with a nationally representative sample. PUBLIC HEALTH RELEVANCE: This study will inform programs that target adolescent alcohol use by providing empirical findings about the long-term effects of adolescent friendship networks on future alcohol use. The project will highlight the appropriate mechanisms to target for prevention and health - peer selection and/or peer influence - and identify groups most susceptible to each mechanism.